Perceptual Color Maps
A color map should map data to color so that equal data steps look equal, and so that structure comes from the data, not the palette.
The problem with rainbow
The traditional rainbow (jet) color map is perceptually non-uniform: it has bright bands and dull stretches, so equal data intervals look unequal. It creates false boundaries at the yellow and cyan transitions and disappears in grayscale. Despite familiarity, it misleads and should be avoided for quantitative data.
Perceptual uniformity
A perceptually uniform map is built in a color space that models human vision (such as CIELAB) so that a fixed step in data produces a roughly constant perceived change. Maps like viridis and its relatives achieve this, are colorblind-friendly, and remain interpretable when printed in grayscale.
- Sequential maps for ordered data from low to high.
- Diverging maps for data with a meaningful midpoint, with a neutral center.
- Categorical palettes for unordered classes, not a continuous ramp.
Match the map to the data
Using a diverging map for one-sided data invents a center that is not there; using a sequential map for signed data hides the sign. Choose the family from the data's structure, then verify it under simulated color-vision deficiency and in grayscale.
Lightness carries the signal
Because lightness is the most reliably perceived channel, a good map varies monotonically in lightness. That is why viridis reads correctly even where hue perception fails.
Kronos use
Kronos figures use perceptually uniform, colorblind-safe maps and avoid rainbow palettes so that structure seen in a field is real and not an artifact of color.